AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Meta-Mining Discriminative Samples for Kinship Verification

Wanhua Li, Shiwei Wang, Jiwen Lu, Jianjiang Feng, Jie Zhou

CVPR 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Dense features and foundation priors · 29 citations (OpenAlex)

Check your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Apply DSMM meta-mining instead of the strongest listed baselineKinFaceW-I verification · KinFaceW-Imean accuracy (%): +3.7%Improved
Apply DSMM meta-mining instead of the strongest listed baselineKinFaceW-II verification · KinFaceW-IImean accuracy (%): +4.6%Improved
Apply DSMM meta-mining instead of the strongest listed baselineTSKinFace verification · TSKinFacemean accuracy (%): +4.9%Improved
Apply DSMM meta-mining instead of the strongest listed baselineCornell KinFace verification · Cornell KinFacemean accuracy (%): +8.8%Improved
Add the DSMM meta-miner to the relation model (ablation)KinFaceW-I relation model · KinFaceW-Imean accuracy (%): 78.1 -> 82.4Improved
Add the DSMM meta-miner to the relation model (ablation)KinFaceW-II relation model · KinFaceW-IImean accuracy (%): 90.4 -> 93.0Improved
Use focal loss instead of meta-learned sample weighting (ablation)KinFaceW-I and KinFaceW-II training · KinFaceW-I; KinFaceW-IImean accuracy (%): 80.9 -> 74.6; 91.8 -> 80.5Hurt
Use unbalanced batches with fixed negative weights instead of balanced-batch resampling (ablation)KinFaceW-I and KinFaceW-II training · KinFaceW-I; KinFaceW-IImean accuracy (%): 80.9 -> 80.9; 91.8 -> 92.5Mixed
Set C to 4 rather than C=1, 2, or 8 (ablation)DSMM training · KinFaceW-Imean accuracy (%): 81.8; 80.6; 82.4; 79.8Mixed
Set C to 2 rather than C=1, 3, or 4 (ablation)DSMM training · KinFaceW-IImean accuracy (%): 92.5; 93.0; 91.3; 90.5Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy